AMD's latest high-end workstation graphics accelerator packs a huge punch, but is it best suited for 4K production? We check it out to bring you the low-down…
Product: Sapphire AMD FirePro W9100
MSRP: $3,999 US (approx. £2,410)
Release date: 7 April 2014
Star Rating: 4 / 5
It has been a while since AMD launched the FirePro W-series of professional 3D accelerators in August 2012. That's quite a long time in the graphics card business, with new GPU generations arriving on an annual basis or even more frequently.
launched its K-series Quadros
almost exactly a year ago, based on its Kepler generation, and its Maxwell generation has already rolled out for consumer platforms, so Quadro iterations will be imminent.
The AMD FirePro W9100 has a 290W maximum draw, so requires the full complement of PCI Express power connections.
AMD needed something to head it off at the pass, and what the company came up with was the Graphics Core Next 1.1 refresh to its Graphics Core Next 1.0 GPUs, the first professionally oriented example of which is the FirePro W9100
we have on review here.
More stream processors, much more memory
The FirePro W9100's numbering implies that it's merely a small improvement over the W9000, but the specification really doesn't bear that out. Where the W9000 offers 2,048 Stream Processors and 128 texture units, the W9100 sports 2,816 Stream Processors and 176 texture units, increases of 37.5 per cent in both cases.
This won't necessarily lead to a 37.5-percent boost in performance, as the clock speed has some influence too, and the relationship is not a linear one anyway due to the overheads associated with parallel thread processing. The core clock has actually been reduced slightly from 975MHz to 930MHz, although this only amounts to a 5-percent drop.
"The most significant factor is that this graphics card sports a staggering 16GB of GDDR5 memory, where the W9000 only offers 6GB, and even NVIDIA's Quadro K6000 only has 12GB"
The GDDR5 memory also runs at a slightly slower 5GHz, compared to the W9000's 5.5GHz, although here again this is a mere 10-percent reduction. However, this should also be weighed against the 512-bit memory bus of the W9100 compared to the 384-bit memory bus of the W9000. The wider bus means the W9100 can provide 320GB per second memory bandwidth compared to the W9000's 264GB per second.
But the most significant factor is that this graphics card sports a staggering 16GB of GDDR5 memory, where the W9000 only offers 6GB, and even NVIDIA's Quadro K6000 only has 12GB. The W9100 is aimed at the rapidly increasing interest in 4K production.
The W9100 name may only be an increment over the W9000, but there is a lot more graphics power contained within.
Power consumptions and connectivity
Amazingly, while the transistor count for all the extra Stream Processors has ballooned from 4.31 billion to 6.2 billion, the thermal design power has only increased by 16W from 274W to 290W. However, although the W9100 has a considerably more powerful GPU than the W9000, the latter was the significant design change, with the switch from VLIW SIMD to RISC SIMD, which improves GPGPU performance.
"The W9100 can provide truly massive arrays of screens, with 3D acceleration available on all of them at once. There's a sync connection for stereoscopy as well"
The updated GPU therefore has only minor enhancements to its graphics feature abilities, with support for DirectX 11.2, OpenGL 4.3, and OpenCL 1.2 – all a 0.1 dot increment higher. However, Shader Model support remains at 5.0, since there is nothing beyond this just yet.
The W9100 provides the same complement of 6 mini DisplayPorts, each one with support for version 1.2 of the standard to drive up to 4K resolutions (4,096 x 2,160). So the W9100 can provide truly massive arrays of screens, with 3D acceleration available on all of them at once. There's a sync connection for stereoscopy as well.
The Sapphire AMD FirePro W9100 is a massive, full-length card requiring 2 slots, but it packs in 16GB of frame buffer.
For testing, the Sapphire AMD FirePro W9100 was supplied to me installed in a SCAN 3XS workstation
. This workstation boasts a 3.4GHz Core i7 4930K processor permanently clocked to 4.4GHz, supported by 32GB of 2133MHz DDR3 SDRAM. This is a perfect platform for 3D modeling, and 3dtotal will be providing a full analysis of the whole system in a future review (stay tuned).
My first test was the now venerable Maxon CINEBENCH R11.5
, which has recently been superseded by the R15 update
. Since I have a lot of historical data on R11.5, the W9100's score of 104.88 in the OpenGL portion of the test can be compared to many previous systems. This is in fact the highest result I have ever seen in this test, although CINEBENCH favors the Sapphire AMD FirePro architecture over NVIDIA Quadros. The result in the OpenGL portion of CINEBENCH R15 is 143.38, which I don't have many systems to compare with, but was also the highest I have seen so far.
I then turned to the industry-standard professional 3D benchmark suite, SPECviewperf
. I ran both version 11 and the recently released version 12. As with CINEBENCH R15, I don't have many comparisons for SPECviewperf 12 yet, but SPECviewperf 11 has been around since 2010 so I have plenty to compare here. The results are generally very good, but also somewhat of a mixed bag. The score of 61.54 in the catia-03 viewset, which is based on Dassault Systemes'
CATIA V5 19 and V6 R2009, is good but not up with systems using NVIDIA's Quadro K4000 or above. The result of 89.58 in the ensight-04 test is more impressive, and the highest we have seen in this test, which simulates performance in CEI's Ensight 8.2.
For 3D animators, the results of 83.31 in the lightwave-01 viewset and 92.49 in the maya-03 viewset will be the most significant. These are great scores, however NVIDIA's Quadro K5000 offers greater performance. On the other hand, product designers will be pleased to hear that the result of 76.67 in the sw-02 viewset, based on SolidWorks 2004 from Dassault Systemes, is the highest I have seen.
As already mentioned, I have few comparisons for SPECviewperf 12, which came out in December 2013. Looking at published scores for this benchmark, the results are extremely good across the board, beating all of the currently published scores.
The Sapphire AMD FirePro W9100 is clearly a very powerful card, and won't disappoint whatever applications you run. However, as always the decision whether to choose this card, a cheaper Sapphire AMD FirePro, or one of their NVIDIA competitors really depends on what software you run.
So if you are running a 4K production workflow and need all the frame buffer you can get, the Sapphire AMD FirePro W9100 is an amazing professional 3D accelerator. Otherwise, you may find more cost-effective options out there.
Shader Model 5.0
2,816 Stream Processors
16GB GDDR5 RAM
6 x Mini DisplayPort
4,096 x 2,160 resolution (DisplayPort 1.2)
290W max power
Windows XP, Vista, 7, and 8
32-bit or 64-bit Linux
Maxon Cinebench R11.5 OpenGL: 104.88
Maxon Cinebench R15 OpenGL: 143.38
Visit the official AMD website
Learn more about FirePro graphics cards
Read more reviews by 3dtotal